Functionally identical to the TI-57 LCD programmable calculator, the TI-57II offers one major improvement: it has a usable keyboard! The first liquid crystal successor to the venerable TI-57 had that horrible keyboard, used across the board on many Texas Instrument handheld calculators, that made these machines a royal pain to use. If you ever come across a Texas Instruments calculator from the mid-1980s that has a keyboard that is incredibly stiff and prone to bounce, it's not an age problem: these keyboards were just as bad when new. How a company like Texas Instruments, known for its innovative handheld keyboard technology since the dawn of the calculator age, could produce such monstrosities remains a mystery.

But, they eventually fixed the problem: several TI calculators were released in updated versions with a keyboard that was, in contrast, a pleasure to use: bounce-free keys that operate with a light, uniform touch even today, on machines 10 to 15 years of age.

Sadly, TI chose not to improve other features of the LCD version of the TI-57: most notably, they chose not to add more memory to this calculator. Either 49 programming steps or 7 memories makes the device unnecessarily limited, and features such as labels, subroutines, and a loop instruction almost feel like a waste.

Just how difficult is it to squeeze algorithms into the limited memory of this device? My Gamma function example demonstrates this very well. Were it not for an otherwise excellent, fully merged programming model, the following program would not even be possible on this device; as things are, I managed to fit it just barely.

What this program computes is the logarithm of the Gamma function using Stirling's formula. An extra loop construct takes care of small or negative arguments where the approximation would be inaccurate. As a result, the value of the Gamma function is computed with typically 8+ digits of precision across the entire domain. For this, each and every program step and memory location that this calculator could offer was needed! The algorithm is reasonably speedy, despite the fact that the calculator itself is blindingly slow.

 61.00     00   STO 0
 51        01   x-t
 09        02   9
 51        03   x-t
 23.00     04   LBL 0
 27        05   x>=t
 22.01     06   GTO 1
 85        07   +
 01        08   1
 95        09   =
 61.65.00  10   STO×0
 22.00     11   GTO 0
 23.01     12   LBL 1
 85        13   +
 81.00     14   EXC 0
 32        15   lnx
 75        16   -
 71.00     17   RCL 0
 65        18   ×
 32        19   lnx
 75        20   -
 53        21   (
 02        22   2
 65        23   ×
 49        24   π
 65        25   ×
 71.00     26   RCL 0
 54        27   )
 35        28   √
 32        29   lnx
 85        30   +
 53        31   (
 06        32   6
 33        33   1/x
 55        34   /
 71.00     35   RCL 0
 34        36   x2
 75        37   -
 05        38   5
 54        39   )
 55        40   /
 06        41   6
 00        42   0
 55        43   /
 71.00     44   RCL 0
 95        45   =
 94        46   +/-
 13        47   R/S